Bulletin of the World Health Organization

Incidence of influenza-like illness and severe acute respiratory infection during three influenza seasons in Bangladesh, 2008–2010

Eduardo Azziz-Baumgartner a, ASM Alamgir b, Mustafizur Rahman a, Nusrat Homaira a, Badrul Munir Sohel a, MA Yushuf Sharker a, Rashid Uz Zaman a, Jacob Dee a, Emily S Gurley a, Abdullah Al Mamun a, Syeda Mah-E-Muneer a, Alicia M Fry c, Marc-Alain Widdowson c, Joseph Bresee c, Stephen Lindstrom c, Tasnim Azim a, Abdullah Brooks a, Goutam Podder a, M Jahangir Hossain a, Mahmudur Rahman b & Stephen P Luby c

a. International Centre for Diarrhoeal Disease Research, 68, Shaheed Tajuddin Ahmed, Sarani, Mohakhali, Dhaka 1212, Bangladesh.
b. Institute of Epidemiology, Disease Control & Research, Dhaka, Bangladesh.
c. Centers for Disease Control and Prevention, Atlanta, United States of America.

Correspondence to Eduardo Azziz-Baumgartner (e-mail: eha9@cdc.gov).

(Submitted: 05 May 2011 – Revised version received: 12 August 2011 – Accepted: 22 August 2011 – Published online: 04 October 2011.)

Bulletin of the World Health Organization 2012;90:12-19. doi: 10.2471/BLT.11.090209

Introduction

Influenza is believed to be an important cause of morbidity and mortality worldwide but data about its burden in low-income tropical countries like Bangladesh are rarely available. Data on incidence as well as on risk factors are valuable in prioritizing influenza prevention and control efforts in the light of competing health interventions.1,2 Data documenting a high incidence of influenza in the United States of America and Europe has helped guide investments in influenza control and prevention in high-income countries.37 However, there is little information about the incidence of influenza in low-income tropical countries.8,9

According to health authorities, the 2009 pandemic influenza A (H1N1) virus caused approximately 6000 deaths in Bangladesh10 and cost Dhaka 6.1 million United States dollars (US$) in direct medical costs to the patient.11 Infrequent hand washing and poor respiratory hygiene (e.g. covering mouth when coughing),12 limited access to care, lack of awareness of antiviral treatment and its availability, a huge shortage of influenza vaccines early in the pandemic, and a high prevalence of malnutrition13 may have worsened the pandemic in low-income countries such as Bangladesh.14 In general, the rate of hospitalization associated with seasonal and pandemic influenza in low-income tropical and subtropical countries remains largely unknown.

In 2004, the International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B) initiated influenza surveillance in a population-based site among children aged less than 5 years in urban Dhaka. The incidence of laboratory-confirmed influenza infection in the site during 2004–2007 was estimated at 10.2 per 100 person–years among children aged less than 5 years seeking care at ambulatory clinics.15 Beginning in April 2007, the Government of Bangladesh and ICDDR,B initiated national hospital-based influenza surveillance with a network of 12 surveillance hospitals throughout the country to provide nationally-representative data from all age groups. Although this national surveillance system helped to establish the seasonality of influenza, which typically occurs during the monsoon season from May to September,16 the incidence of influenza could not be estimated because people in hospital catchment areas sought care not just in hospitals but in other facilities as well and data on the population at risk, needed for the denominator, were not available. In the present study, we used health utilization patterns in the catchment areas of four selected sentinel hospitals to estimate the incidence of both outpatient and hospital visits for seasonal and pandemic influenza illness during three influenza endemic periods during 2008–2010.

Methods

Enhanced hospital-based surveillance

Investigators selected one private and three government hospitals, each in a different region of Bangladesh. The four hospitals, which were participating in the 12-hospital Government of Bangladesh/ICDDR,B national hospital-based influenza surveillance programme, were Jahurul Islam Medical College Hospital in Kishorgonj; Comilla Medical College Hospital in Comilla; Shahid Ziaur Rahman Medical College Hospital in Bogra, and Sher-e-Bangla Medical College Hospital in Barisal. Beginning in May 2008, surveillance physicians identified patients with influenza-like illness (ILI, defined as sudden onset of subjective fever, cough or sore throat) in the outpatient departments at sentinel sites during two randomly selected days every month. Investigators only selected patients who resided in the subdistricts belonging to the hospital’s main catchment areas (i.e. where the majority of hospital patrons lived), as determined by reviewing hospital log books. Beginning in January 2009, also during two randomly selected days a month, surveillance physicians identified all cases of severe acute respiratory illness (SARI) among children less than 5 years old in sentinel site paediatric wards. To identify these children, physicians applied the definition of severe pneumonia:17 cough or difficulty breathing and at least one danger sign (i.e. chest indrawing, stridor while calm, history of convulsions, inability to drink, lethargy or unconsciousness, and/or intractable vomiting).18 Beginning in January 2009, surveillance physicians identified all patients aged 5 years or older admitted to medical wards who met the SARI case definition of subjective fever within the past 21 days and cough or sore throat.

Each physician collected clinical information using a structured questionnaire and obtained a nasal and throat swab from every eligible patient who presented within 7 days of symptom onset. Each participant’s nasal and throat swabs were pooled in viral transport media and transported to the ICDDR,B laboratory in a cool box at 2–8 °C. Specimens were stored at ≤ −70 °C until tested for the presence of influenza viruses by real-time reverse transcription polymerase chain reaction.19 Influenza A viruses were subtyped using H1, H3 and H5 primers provided by the Influenza Division at the United States Centers for Disease Control and Prevention.20 These tests were anticipated to have a sensitivity of 95% and specificity of 98%.21

Hospital catchment assessment

From 1 July 2008 and 6 January, ICDDR,B conducted a health care utilization survey of the four hospital catchment areas to determine the size of the population served by each hospital. To calculate the proportion of people with sudden fever, cough and sore throat (ILI) who visited or were admitted to sentinel sites out of all people with similar symptoms in the community, we estimated that we needed a sample consisting of 282 individuals with respiratory illness for the two-month period before the survey for a design effect of 2.5. To identify these 282 individuals, we needed to enrol approximately 32 households from each of 30 unions (i.e. an administrative unit of Bangladesh with about 28 000 inhabitants) at every one of the 4 sentinel sites, if we assumed that each household had a mean of 5 dwellers, that 15% of household members had had a respiratory infection during the 2 months preceding the survey, and that 10% of these individuals had sought care in a catchment area hospital (i.e. 288 = 32 × 30 × 5 × 4 × 0.15 × 0.10). We randomly sampled 30 unions from each hospital catchment area using a sampling strategy that accounted for the number of people living in each union (i.e. population proportionate to size sampling). Investigators programmed a spreadsheet to generate random spatial coordinates within the boundaries of the selected unions. Using global positioning systems, teams travelled to the randomly selected coordinates and approached the nearest dwelling for enrolment.

An interviewer in each site collected information on demographics (e.g. age, sex), pre-existing conditions (e.g. diagnosed lung or heart disease), history of respiratory illness (e.g. sudden onset fever, cough, sore throat, shortness of breath during the preceding week, 2 months and 1 year), and care-seeking behaviour (e.g. ambulatory care sought at sentinel site 2 months or less before the interview or hospitalization 1 year or less before the interview) for all household members. The interviewer then skipped the next two closest households and approached the third dwelling for enrolment until information was collected from 32 households in each of the 30 unions.

Data analysis

The investigators used census data to estimate the size of the age-specific populations living in the catchment areas during 2001 and applied a 1.5% annual growth rate to estimate their population in 2008 and 2009.22 The annual growth rate was estimated using crude birth rates and the national crude death rate, on the assumption that net migration was minimal.23 We estimated the rate of influenza among people with ILI by dividing the number of patients who presented to health services at sentinel sites for ILI and whose samples tested positive for influenza viruses by the census population of each catchment area and adjusting the results for the proportion of days during which active surveillance for ILI had been conducted (i.e. two days a month). We also adjusted for the proportion of the census population seeking care from other health-care providers (as per the health care utilization survey).

To estimate the rates of SARI associated with influenza (henceforth referred to as influenza-associated SARI) during the endemic period16 we multiplied the proportion of persons seeking care at an ambulatory clinic who were later admitted to wards (i.e. h/o, where h represents ILI patients admitted to hospital from the outpatient department with influenza, and o represents ILI patients in the outpatient department with confirmed influenza) by the estimated rate of ILI associated with influenza (henceforth referred to as influenza-associated ILI). We estimated the rate of influenza-associated ILI in the community by dividing the number of persons with samples positive for influenza viruses seeking care for ILI at the sentinel site (i.e. o) by the census population of the catchment area and adjusting (i.e. dividing) by the proportion of days during which active surveillance for ILI was conducted (i.e. two days a month during the year, or 2 × 12/365). We also adjusted (i.e. divided) by the proportion of the census population seeking care at facilities other than sentinel hospitals (i.e. s/c where s represents the number of patients with ILI who sought care at the sentinel site during the 2 months that preceded the health care utilization survey, and c represents the number of patients living in the catchment area who sought care from any provider for subjective fever, cough or sore throat during 2 months before the survey.

We used nonparametric bootstrapping methods to estimate the 95% confidence intervals (CIs) of the incidence rates for ILI and influenza-associated SARI .24 Investigators calculated the rates of influenza-associated SARI only for 2009 and 2010, when the hospitals started to routinely identify these patients. To estimate the annual number of children aged < 5 years with influenza-associated hospitalizations in Bangladesh, we multiplied the incidence of influenza-associated SARI calculated from the four catchment areas by the population of children aged < 5 years in Bangladesh during 2009.22 Similarly, to estimate the number of persons aged ≥ 5 years with influenza-associated hospitalizations, investigators multiplied the incidence of influenza-associated SARI among those aged ≥ 5 years calculated from the four catchment areas by the populations of persons of that age group during 2009. We performed Wilcoxon rank sum, χ2 and Fisher’s exact tests where appropriate.

Human subjects

We requested written informed consent from patients at inpatient and outpatient departments in sentinel sites and from adult heads of household before enrolling participants in the surveillance system or in the household survey. The protocol was reviewed and approved by ICDDR,B and by the institutional review board of the Centers for Disease Control and Prevention in Atlanta, United States. During the pandemic, the team encouraged hospital staff to administer olsetamivir to all SARI and ILI patients who were aged less than 5 years or 65 years or over, who were pregnant, or who had any of the following: diabetes; chronic heart, lung or liver disease; asthma; neurologic or neuromuscular, haematologic or metabolic disorders; immune suppression; cancer or obesity, in accordance with Government of Bangladesh national treatment guidelines.

Results

Hospital-based influenza surveillance

From May 2008 to September 2010, surveillance physicians identified SARI in 358 patients aged less than 5 years and 495 aged 5 years or more; they also identified 2208 persons with ILI (Table 1). Most patients lived close to a sentinel hospital. The median age of 399 children less than 5 years old who were admitted to the paediatric ward with SARI was 4 months (range: 2–9). Only 101 (25%) of these children were female, compared with 413 (42%) of the 976 children less than 5 years old who had ILI (P < 0.001). Medical ward inpatients were significantly older than outpatients (median age: 37 and 20 years, respectively; P < 0.0001). The proportion of females aged 5 years or older admitted to hospital wards with SARI (168 [34%] of 495 with SARI) was significantly lower (P < 0.001) than the proportion that had ILI (643 [52%] of 1232 with ILI).

Clinical characteristics

ILI and SARI patients presented for care a median of 4 days after fever onset. Of 358 SARI patients aged less than 5 years, 314 (88%) had had fever. Of these 358 patients, 17 (5%) had known pre-existing lung disease, compared with 9 (1%) of 970 ILI patients of the same age (P < 0.001). Similarly, of 495 SARI patients aged 5 years or older, 89 (1%) had known pre-existing lung disease, such as chronic obstructive pulmonary disease or asthma, compared with 39 (3%) of 1216 ILI patients the same age (P < 0.001).

Patients with positive samples

Respiratory samples from 7 (9%) of 82 SARI patients aged less than 5 years, 1 (5%) of 21 SARI patients aged 5 years or older, and 93 (16%) of 586 ILI patients of all ages tested positive for influenza viruses during 2008. This was in contrast to respiratory samples from 5 (5%) of 100 SARI patients aged less than 5 years, 55 (23%) of 236 SARI patients aged 5 years or older, and 115 (12%) of 944 ILI patients tested positive for influenza during 2009; and in contrast to 4 (2%) of 181 SARI patients aged less than 5 years, 54 (23%) of 238 SARI patients aged 5 years or older, and 158 (23%) of 678 ILI patients during 2010. Among 631 patients whose samples tested positive for influenza viruses, 325 (64%) tested positive for influenza A (i.e. 74 [12%] for seasonal influenza type A H1, 164 [26%] for type A H3, and 174 [28%] for type A pH1); 162 (32%) tested positive for influenza B, and 5 (1%) tested positive for both influenza A and B (Fig. 1). During 2008, 78% of patients infected with seasonal influenza viruses presented to the hospitals from May to October and peak activity occurred in July.

Fig. 1. Patients with severe acute respiratory infection and influenza-like illness who were confirmed to have influenza, Bangladesh, 2008–2010
Fig. 1. Patients with severe acute respiratory infection and influenza-like illness who were confirmed to have influenza, Bangladesh, 2008–2010

Health care utilization survey

From July 2008 to January 2009, field workers interviewed participants from 3840 households and obtained information on health care utilization for 17 495 household members. The median age of household members was 22 years and 53% were female. Participants identified 2435 (13.9%) household members who had had fever and cough or a sore throat (ILI) the week before the interview (Table 2). Participants reported that 484 children aged less than 5 years had sought care for ILI from some type of provider during the 2 months before the interview, and that 16 (3%) of these children had sought care at the sentinel hospitals. Similarly, 1950 people aged 5 years or older sought care for ILI from some type of provider during the 2 months before the interview, and 84 (4%) of them had done so at the sentinel hospitals. Respondents also reported that 165 (0.9%) of the 17 495 household members had been admitted to hospital for an illness that had started with fever and cough or a sore throat (i.e. SARI) during the year before the interview. Of these 165 SARI patients, 61 (37%) had been admitted to a sentinel hospital.

Hospitalizations and clinic visits for influenza

From 2008 to 2010, seven SARI patients aged less than 5 years whose samples tested positive for influenza viruses were admitted from the paediatric outpatient department (i.e. 3 in 2008, 3 in 2009 and 1 in 2010). Similarly, 17 SARI cases whose samples tested positive for influenza viruses were admitted from the adult outpatient department (i.e. 8 in 2009 and 9 in 2010). Based on these findings, the estimated incidence of SARI associated with influenza among children aged less than 5 years was 6.7 cases per 1000 person­–years (py) (95% CI: 0–18.3) during the 2008 season, 4.4 per 1000 py during the 2009 season (95% CI: 0–13.4) and again 6.5 per 1000 py (95% CI: 0–8.3) during the 2010 season. The incidence of SARI among people aged 5 years or over was 1.1 cases per 10 000 py (95% CI: 0.4–2.0) during 2009 and 1.3 per 10 000 py (95% CI: 0.5–2.2) during 2010.

Multiplying our average SARI incidence (5.9 per 1000 py) by the population of children aged less than 5 years in Bangladesh during 2010 (i.e. 19 331 302),22 we estimated that annually 113 000 children in this age group seek hospital care for influenza. Multiplying the average SARI incidence (i.e. 1.2 cases per 10 000 py) by the population of persons aged 5 years or older in Bangladesh during 2010 (i.e. 132 920 875),22 we also estimated that 16 000 people in this age group annually seek hospital care for influenza.

The incidence of medically attended, laboratory-confirmed seasonal influenza among outpatients with ILI was 10 cases per 100 py (95% CI: 8–14) during the 2008 season, 6.6 per 100 py (95% CI: 5–9) during the 2009 season and 17 per 100 py (95% CI: 13–22) during the 2010 season (Table 3). Similarly, the incidence of ambulatory ILI among patients infected with the pandemic (H1N1) virus was 4.5 cases per 100 py (95% CI: 3–6) during 2009.

Discussion

In our study, the rate of clinic visits for influenza during non-pandemic years was similar to that observed in other sites in Bangladesh and southern Asia.25 In an urban community-based surveillance site in Bangladesh, the incidence of influenza among children aged less than 5 years with acute respiratory infection was 10 per 100 py.15 The incidence of ILI (10 per 100 py in 2008) is also similar to that observed in rural India in children aged less than 3 years with acute respiratory infection and influenza (14 per 100 py).26 Our ILI estimates, however, are an order of magnitude higher than in Thailand (14 per 1000 py).27 In Thailand they sampled ILI patients with measured fever (≥ 38 °C), whereas in Bangladesh we sampled ILI patients with subjective fever, an approach likely to capture a greater number of influenza patients. In addition, in Thailand they used health care utilization data for cases with pneumonia to determine the effective catchment population of their sites and this may have underestimated the percentage of people seeking care for ILI outside their surveillance facilities.

This manuscript has several important limitations. We assumed that we could accurately determine the average proportion of samples that tested positive for influenza viruses by randomly sampling patients two days a month, since positivity changed gradually over the 6-month period of influenza endemicity in Bangladesh. In addition, health care utilization surveys were conducted sequentially during 6 months in each catchment area before the pandemic and it is possible that health care seeking patterns documented in the survey differed from those observed during the pandemic. Nevertheless, health care utilization patterns are limited by the capacity of the health-care system to treat patients in Bangladesh, where 11 patients are hospitalized for every 10 hospital beds.28 Incidence was estimated from four sentinel sites that may not have been representative of all of Bangladesh, particularly its urban areas, which are densely populated and more prosperous than rural areas and which represent approximately 25% of the country’s population. We were unable to reliably differentiate ILI patients with mild disease from those with severe respiratory illness in the community who were never hospitalized. This may have led us to underestimate SARI incidence among catchment area residents. Nevertheless, our estimated incidence of influenza was similar to the incidence detected through active community-based surveillance in urban Dhaka (i.e. 10.2 per 100 py).15

Conclusion

Influenza illness in Bangladesh is often associated with health care seeking in the outpatient setting. In addition, children aged less than 5 years seem disproportionately affected by influenza-associated hospitalization, with rates that were higher than those seen among Bangladeshi adults or young children in the United States.29 Our incidence figures for outpatient visits and hospitalizations for influenza suggest that influenza control strategies would be beneficial in countries like Bangladesh.


Acknowledgements

ICDDR,B acknowledges with gratitude the commitment of the Centers for Disease Control and Prevention to its research. We would also acknowledge the contributions of Dorothy Southern, Jo Grzelinska, Peter Kim Streatfield, Salah Uddin Khan, Sk. Masum Billah, Nancy Cox, Carolyn Bridges, Ann Moen, Karen Siener, Sonja Olsen, Kimberly Lindblade, Paul Gargiullo and the surveillance staff. The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the Centers for Disease Control and Prevention or the institutions with which the authors are affiliated.

Funding:

This study was funded by the United States Centers for Disease Control and Prevention; grant number U01/CI000628-02.

Competing interests:

None declared.

References

Share